Affective computing of multi-type urban public spaces to analyze emotional quality using ensemble learning-based...

Area of Science:

• Urban Planning
• Affective Computing
• Human-Computer Interaction

Background:

• Urban public space quality significantly impacts user emotional responses.
• Emotional data offers a quantifiable metric for evaluating space quality.
• Affective computing provides a framework for measuring emotions in public spaces.

Purpose of the Study:

• To propose and validate a method for evaluating multi-type urban public space quality using physiological signals and ensemble learning.
• To assess the effectiveness of affective computing in urban space quality assessment.
• To provide evidence-based support for urban space renewal initiatives.

Main Methods:

• Experiments were conducted in eight public spaces across five types.
• Participants' physiological signals and self-reported emotions were collected.
• Binary, ternary, and quinary classification models were developed using ensemble learning.

Main Results:

• Binary and ternary classification models achieved high accuracies (up to 92.59% and 91.07% respectively).
• External validation confirmed model effectiveness for urban space quality evaluation (up to 80.90% accuracy).
• Ensemble learning improved accuracy by an average of 7.59% compared to single classifiers.

Conclusions:

• The proposed method, utilizing physiological signals and ensemble learning, shows promise for evaluating urban public space quality.
• The study highlights the potential of affective computing in evidence-based urban renewal.
• Optimizing feature selection and addressing data imbalance are key to enhancing evaluation accuracy.